A Novel Method for Screening Children with Isolated Bicuspid Aortic Valve

Area of Science:

• Cardiology
• Biomedical Engineering
• Artificial Intelligence

Background:

• Congenital heart malformations, such as isolated bicuspid aortic valve (BAV), pose significant cardiovascular risks.
• Early screening and diagnosis are crucial for managing BAV and preventing serious complications.
• Current diagnostic methods may face challenges with non-stationary heart sound signals and background noise.

Purpose of the Study:

• To introduce a novel signal processing method, the statistical time growing neural network (STGNN), for robust heart sound classification.
• To develop an automated tool for screening children with isolated BAV using advanced signal processing and machine learning.
• To evaluate the performance of the STGNN against existing methods in classifying pediatric heart sounds.

Main Methods:

• Development of the statistical time growing neural network (STGNN) by integrating supervised and unsupervised statistical learning.
• Application of preprocessing and segmentation techniques to heart sound signals.
• Comparative analysis of STGNN, conventional time growing neural network (CTGNN), and conventional support vector machine (CSVM) using balanced repeated random sub-sampling.
• Inclusion of 22 children with BAV and 28 healthy children in the study.

Main Results:

• The STGNN achieved an average accuracy of 87.4% and sensitivity of 86.5%.
• STGNN outperformed CTGNN (81.8% accuracy, 83.4% sensitivity) and CSVM (72.9% accuracy, 66.8% sensitivity).
• STGNN demonstrated superior performance and enhanced immunity to background noise compared to CTGNN and CSVM.

Conclusions:

• The STGNN offers a robust and accurate method for heart sound signal classification, particularly for screening pediatric BAV.
• The developed automated tool shows promise for implementation in primary healthcare settings to improve early detection of congenital heart defects.
• The STGNN's resilience to noise makes it a valuable advancement in non-invasive cardiac diagnostics.